Multi-label Learning for Chronic Disease Prognosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical examination systems lack effective means for predicting the occurrence of chronic diseases in the future, particularly failing to account for concurrent correlations between multiple chronic diseases, leading to decreased prediction accuracy and limited prognostic capabilities.
Innovation Solution
A multi-label learning system is developed, comprising a data acquisition, preprocessing, and prediction module, which constructs a multilayer neural network using standardized feature vectors and loss functions to predict the occurrence of chronic diseases within 1 to 5 years, incorporating time sequence information and logical relations between diseases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional single-label machine learning method is used to predict various diseases, then the prediction can be performed, but the concurrent correlation between different chronic diseases cannot be extracted, resulting in decreased prediction accuracy
Solution Approach 1:
The patent segments the prediction task into multiple independent binary classification problems, where each chronic disease is predicted separately using dedicated neural network branches. This allows the system to handle multiple diseases concurrently while maintaining the ability to capture correlations between them through shared feature extraction layers, thereby resolving the contradiction between prediction accuracy and the ability to handle concurrent diseases.
Solution Approach 2:
The patent implements a multi-label learning framework where a single system performs prediction for multiple chronic diseases simultaneously. The shared neural network structure processes examination data once and generates predictions for all target diseases, making the system versatile for handling concurrent diseases while maintaining high prediction accuracy through specialized disease-specific branches.
2Loss of time
If current medical examination system determines disease presence at current time, then immediate diagnosis is achieved, but prognosis of occurrence of chronic diseases in the next several years cannot be performed
Solution Approach 1:
The patent applies preliminary action by training the multi-label learning model on historical examination data with known disease outcomes. The system learns temporal patterns and risk factors that precede disease development, enabling it to perform prospective prognosis predictions for future disease occurrence based on current examination results, thus achieving both immediate diagnosis and future prediction capabilities.
3Adaptability or versatility
If multi-label machine learning is used for diagnosis at current time node, then concurrent diseases can be identified, but prognostics on occurrence of future chronic diseases cannot be performed
Solution Approach 1:
The patent extends the prediction problem into a new dimension by incorporating time as a critical factor. The system not only predicts which diseases are present now but also estimates the probability and timing of future disease occurrence. This temporal dimension transformation allows the multi-label framework to simultaneously provide cross-sectional diagnosis and longitudinal prognosis, resolving the contradiction between multi-disease detection and future prediction capabilities.
Data Source
AI summary
Provided is a system for the prognostics of the chronic diseases after the medical examination based on the multi-label learning, including a data acquisition module, a data preprocessing module, a basic predicting model constructing module, and a local predicting module. The data acquisition module is configured to acquire physical examination data of a physical examination user. The basic predicting model constructing module is configured to construct a multi-label learning model for a physical examination scenario. The local predicting module includes a local model training unit and a predicting unit. The local model training unit adjusts the basic predicting model into a local predicting model, and solidifies the local predicting model into the local predicting module. The predicting unit outputs a predicted prognostic index for an occurrence of a plurality of chronic diseases, and finally acquires a future expected occurrence time of the chronic diseases.